Context-Based Features Enhance Protein Secondary Structure Prediction Accuracy

Abstract

We report a new approach of using statistical context-based scores as encoded features to train neural networks to achieve secondary structure prediction accuracy improvement. The context-based scores are pseudo-potentials derived by evaluating statistical, high-order inter-residue interactions, which estimate the favorability of a residue adopting certain secondary structure conformation within its amino acid environment. Encoding these context-based scores as important training and prediction features provides a way to address a long-standing difficulty in neural network-based secondary structure predictions of taking interdependency among secondary structures of neighboring residues into account. Our computational results have shown that the context-based scores are effective features to enhance the prediction accuracy of secondary structure predictions. An overall 7-fold cross-validated Q3 accuracy of 82.74% and Segment Overlap Accuracy (SOV) accuracy of 86.25% are achieved on a set of more than 7987 protein chains with, at most, 25% sequence identity. The Q3 prediction accuracy on benchmarks of CB513, Manesh215, Carugo338, as well as CASP9 protein chains is higher than popularly used secondary structure prediction servers, including Psipred, Profphd, Jpred, Porter (ab initio), and Netsurf. More significant improvement is observed in the SOV accuracy, where more than 4% enhancement is observed, compared to the server with the best SOV accuracy. A Q8 accuracy of >70% (71.5%) is also found in eight-state secondary structure prediction. The majority of the Q3 accuracy improvement is contributed from correctly identifying β-sheets and α-helices. When the context-based scores are incorporated, there are 15.5% more residues predicted with >90% confidence. These high-confidence predictions usually have a rather high accuracy (averagely ~95%). The three- and eight-state prediction servers (SCORPION) implementing our methods are available online.

FLEXc: protein flexibility prediction using context-based statistics, predicted structural features, and sequence information

“…In this work, similar to our previous work employed in DINOSOLVE [29], SCORPION [30], and CASA, we collect statistics of singlets (), doublets (), and triplets () residues at different positions in protein chains in a window of size 7 residues (). These statistics represent approximations of the possibilities of residues adopting certain flexibility states when none, one, or two neighboring residues are considered.…”

“…We use the methods SCORPION [30] and CASA for secondary structure and solvent accessibility predictions, respectively.…”

“…We describe the approaches of extracting statistical scores to measure the favorability of residues’ flexibility in presence of its surrounding neighbors in sequence from a large training dataset based on the mean-field potentials [28]. These approaches were successfully applied in our previous work for predicting protein disulfide bonding [29], secondary structures [30, 31], and solvent accessibility. The basic idea is based on the observation that residues’ flexibility exhibit strong local dependency.…”

A Distributed Tree-based Ensemble Learning Approach for Efficient Structure Prediction of Protein

“…Computational methods for PSSP, mostly based on machine learning methods, can be schematically grouped in the following three families: sequence- based methods; network-based methods [9]; hierarchical ensemble methods [10]. Some methods provided predictions of a relatively small set of functional classes [11], while others considered predictions extended to larger sets, using Support Vector machines [12] , HMM Algorithm, artificial neural networks, Bayesian networks, Decision Trees or methods that combine functional linkage networks with learning machines using a logistic regression model or simple algebraic operators.…”

Prediction of 8-state protein secondary structures by 1D-Inception and BD-LSTM

“…Sometimes it is essential to know protein 3D structures to identify the protein functions at a molecular level. Reliably and accurately predicting protein 3D structure form sequences of proteins is one of the most challenging issues in computational biology [1]. Protein secondary structure prediction is a vital step towards to predict protein tertiary (3D) structure [2], prediction of protein disorder [3], and solvent accessibility prediction [4].…”